Development of a continuous assay for high throughput screening to identify inhibitors of the purine salvage pathway in Plasmodium falciparum

Malaria, an infectious disease caused by protozoan parasites from the genus Plasmodium , represents a serious global health threat. The continued emergence of drug resistant strains has severely decreased current antimalarial drug eﬃcacy and led to a perpetual race for drug discovery. Most protozoan parasites, including Plasmodium spp. , are unable to synthesize purines de novo and instead rely on an essential purine salvage pathway for acquisition of purines from the infected host. Because purines are essential for Plasmodium growth and survival, the enzymes of the purine salvage pathway represent promising targets for drug discovery. Target-based high-throughput screening (HTS) assays traditionally focus on a single target, which severely limits the screening power of this type of approach. To circumvent this limitation, we have reconstituted the purine salvage pathway from Plasmodium falciparum in an assay combining four drug targets. This assay was developed for HTS and optimized to detect partial inhibition of any of the four enzymes in the pathway. Inhibitors of several enzymes in the pathway were identiﬁed in a pilot screen, with several compounds exhibiting eﬀective inhibition when provided in micromolar amounts.


Introduction
Malaria is caused by protozoan parasites from the genus Plasmodium , with the species P. falciparum being the most prominent among the pathogenic species contributing to human morbidity and mortality [1] . While great progress has been made to eradicate malaria in the developed world, the disease continues to present a global health burden with ∼230 million clinical incidences and over 400,000 deaths reported in 2019 alone [2] . The challenges associated with malarial infections are compounded by the fact that nearly all clinical cases of the disease now exhibit either singular, or multi-drug resistance to standard treatments ( e.g. , artemisinin and artemisinin combinatorial therapies). [3][4][5][6][7][8] The need for development of novel therapeutics is of vital importance to ensuring that alternate treatments for malaria are available if existing antimalarials lose their efficacy, which according to current trends seems all but inevitable.
P. falciparum is transmitted to humans through the bite of an infected female Anopheles mosquito. Infection is first established in liver hepatocytes, followed by rapid proliferation and a series of destructive erythrocytic cycles [1] . The liver stage and red blood cell stages represent the asymptomatic and symptomatic stages of infection, re-spectively. While the early liver stage presents an attractive target for drug development, without improved and accessible diagnostic techniques, targeting this stage of infection is challenging due to its asymptomatic nature [9] . Therefore, efforts to inhibit essential processes that occur during the erythrocytic stage of infection is a realistic alternative.
Most protozoan parasites, including P. falciparum , are unable to synthesize purines de novo , and instead rely on purine salvage pathways for acquisition of these nucleic acid building blocks from the host. [10][11][12] Because parasitic cells are rapidly proliferating during the erythrocytic stage of infection, importing purines for DNA and RNA synthesis is essential for growth and survival. In fact, culturing of P. falciparum in serum-free media is only made possible by exogenous supplementation of purines, further emphasizing the importance of the purine salvage pathway in this species [13 , 14] . The direct import of purines is facilitated by nucleoside transporters embedded in the parasitic cellular membrane [15 , 16] . While the nucleoside transporters represent appealing drug targets themselves, the three downstream enzymes of the purine salvage pathway have garnered much attention as the focal points for discovery of novel antimalarial compounds (for review see [17 , 18]  In P. falciparum the cytosolic enzymes that make up the purine salvage pathway are adenosine deaminase (PfADA), purine nucleoside phosphorylase (PfPNP), and hypoxanthine-guanine-xanthine phosphoribosyltransferase (PfHGXPRT) ( Figure 1 A ). Adenosine, as well as various intermediates of the purine salvage pathway, are imported from infected erythrocytes and processed via the pathway to yield inosine monophosphate (IMP), a precursor for all purines required by the parasite. Genetic studies carried out in Plasmodium demonstrate the suitability of the enzymes of the purine salvage pathway as potential drug targets. For instance, two independent genetic screens identified PfADA and pfHGXPRT as essential factors for the asexual blood-stage growth of Plasmodium. [19 , 20] . In addition, P. falciparum PfPNP mutants are unable to thrive in vitro at physiological concentrations of hypoxanthine suggesting that salvage of adenosine is required [21] . Plasmodium yoelii parasites (a rodent-specific species) exhibiting a defective PNP are attenuated in a mouse infection model [22] .
In P. falciparum , the purine salvage pathway serves the additional purpose of eliminating 5'-methylthioadenosine (MTA), a byproduct of the polyamine synthesis pathway ( Figure 1 A ) [23] . In humans, MTA is recycled using a different enzyme, MTA phosphorylase. Polyamines are essential polycations found in all domains of life that are involved in many cellular mechanisms [24 , 25] . They constitute 14% of the plasmodial metabolome, [26 , 27] and are essential for parasite proliferation [27] , [28] Elimination of MTA by the purine salvage pathway is important because this compound is a strong inhibitor of the last step of polyamine synthesis [29] . Therefore, disruption of MTA recycling leads to inhibition of polyamine synthesis in Plasmodium .
The active sites of PfADA and PfPNP are of particular interest for drug development because they differ between the plasmodial proteins and their human orthologs [30] , [31] . Both enzymes exhibit dual substrate specificities; PfADA can utilize adenosine and MTA, and PfPNP can use inosine and 5'-methylthioinosine (MTI). Inhibitors targeting these proteins in the purine salvage pathway have yielded mixed results. Analogs of coformycin, for example, were shown to have high affinity for PfADA and to elicit antiplasmodial activity in RBCs when MTA is added as a source of purines [32] . However, even when supplied in only picomolar amounts, these compounds inhibit the human ADA and therefore have limited utility as a safe treatment option [17 , 33] . Another inhibitor, deoxycoformycin, was shown to cause decreased parasitemia in a primate model, [34] demonstrating the suitability of ADA as a target for therapeutics. In recent years, the transition-state inhibitors called Immucillins have been developed to target both PfPNP and human PNP (for review see [18] ). One promising candidate among this class of inhibitors is DADMe-Immucillin-G, which effectively targets both proteins. This compound causes Plasmodium death by purine starvation in RBCs and completes clearance of parasitemia in primate clinical trials [18 , 35] . Despite its high affinity for the human PNP, DADMe-Immucillin-G exhibits a low level of toxicity. In addition to the clinical data regarding enzymes of the purine salvage pathway acting as drug targets, crystal structures are available for all three of the enzymes from P. falciparum , making rational drug design attainable [30 , 31 , 36] .
In Plasmodium , HGXPRT exhibits relaxed substrate specificity, catalyzing the transfer of the phosphoribosyl moiety of -D-phosphoribosyl pyrophosphate (PRPP) to various nucleobases ( i.e. , hypoxanthine, xanthine, and guanine) to form the respective purine nucleoside monophosphates and inorganic pyrophosphate [37] . In contrast, the human enzyme (encoded by HPRT1 ) exhibits tighter substrate specificity and can only utilize guanine and hypoxanthine as substrates. [38] . Acyclic nucleoside phosphonates were developed to selectively target the plasmodial HGXPRT [36] . Although these inhibitors bind tightly to the Plasmodium enzyme and are able to arrest parasite growth in RBCs, [39] , [40] they lack selectivity [41] and move poorly across cell membranes [42] .
In the present study, we have setup an in vitro assay for the simultaneous screening of potential inhibitors of the three enzymes of the purine salvage pathway of P. falciparum . The pathway was reconstituted in vitro and coupled to the IMP dehydrogenase (IMPDH) from Staphylococcus aureus (SaIMPDH). This latter enzyme produces NADH over time allowing for continuous monitoring of the activity of the entire pathway using spectrofluorimetry ( Figure 1 B ). IMPDH has recently gained a lot of traction as a drug target for treating multi-drug resistant bacterial and fungal infections and therefore itself represents a valuable drug target. [43][44][45][46] The total number of targets screened simultaneously in this assay is consequently brought to four. All substrate and enzyme concentrations were optimized to allow for detection of partial inhibition of any of the four enzymes. The sensitivity and reproducibility of this method, combined with the streamlined approach for multi-target drug discovery, makes for an assay that is amenable to HTS. As proofof-concept, a 384-compound subset of the Asinex BioDesign library was screened, leading to identification of inhibitors of the enzymes PfADA, PfHGXPRT, and SaIMPDH.

Cloning and protein purification
The ADA, PNP , and HGXPRT genes were amplified using cDNA from P. falciparum 3D7 (a gift from Dr. D. Chakrabarty, UCF, Florida). The gene encoding IMPDH (GeneBank ABX28441) was amplified using genomic DNA from S. aureus USA300 (a gift from Dr. A. Cole, UCF, Florida). ORFs were cloned using the FastCloning technique [47] into a derivative of vector pet33b (Novagen) in which the thrombin site has been substituted with a TEV site (see primers listed in Table S1 and Figure S1 for details). Proteins were expressed in the BL21 Rosetta2 strain of E. coli (Novagen) in the presence of 50 mg/L of kanamycin and 30 mg/L of chloramphenicol. Strains expressing the P. falciparum proteins were grown in Luria-Bertani broth at 37°C until they reached an OD equivalent to 0.6. Expression of each protein was carried out using variable conditions as follows. PfADA expression was induced by addition of 0.5 mM isopropyl-D-thiogalactopyranoside (IPTG), and bacteria were harvested after 3 h of incubation at 37°C. PfPNP and PfHGXPRT were induced with 0.1mM IPTG followed by 6 h of incubation at 21°C. SaIMPDH was expressed using an autoinduction medium, [48] and bacteria were harvested after incubation overnight at 37°C. N-terminally His6-tagged proteins were purified using TALON affinity chromatography resin (Clontech) according to the guidelines of the manufacturer. Proteins were stored at − 80°C in a buffer containing 100 mM Tris • HCl (pH 8.0), 100 mM NaCl, 3 mM -mercaptoethanol, and 50% (v/v) glycerol. Purified proteins were analyzed via SDS-PAGE to assess purity ( Figure S2 ) and stored at -80°C.

PfHGXPRT activation
Purified PfHGXPRT was activated according to described procedures [49] and with the following modifications. 50 μM PfHGXPRT was incubated in a mixture containing 100 mM Tris-HCl (pH 8.0), 1 mM PRPP, 12 mM MgCl 2, and 5 mM DTT. An aliquot of the activation mixture was removed at various time intervals and HGXPRT activity was assayed immediately ( Figure S2 ).

K m determination
For determination of K m values, 50 L reactions were assembled in 384-well plates (Nunc TM 384-Well Clear Polystyrene). Fluorescence measurements were taken every 10 -30 s using a Biotek Synergy H1 Microplate Reader with the temperature maintained at 21°C. Reaction mixtures contained 100 mM Tris-HCl (pH 8.0), 100 mM KCl, 12 mM MgCl 2 , and 1 mM dithiothreitol (DTT). The K m of SaIMPDH for NAD + was determined in the presence of 52 nM of enzyme, 2 mM IMP, and varying concentrations of NAD + (0.036 -2 mM). The K m of PfHGX-PRT for PRPP was determined in the presence of 0.6 μM of enzyme, 5 mM hypoxanthine, and varying concentrations of PRPP (31 -500 μM). IMP production was detected by coupling the HGXPRT reaction to that of the IMPDH reaction using 1 μM of SaIMPDH and 1 mM of NAD + . The K m of PfPNP for Pi was determined in the presence of 63 nM of enzyme, varying concentrations of Na 2 HPO 4 (31 -1000 μM), and 0.3 mM of 2-amino-6-mercapto-7-methyl purine ribonucleoside, a nucleoside analog in which the free base is detectable by spectrophotometry ( max = 355 nm) [50] . The K m of PfADA for adenosine was determined as previously described, [51] in the presence of 50 nM of enzyme, 5 mM hypoxanthine, and varying concentrations of adenosine (60 -500 μM).

Reconstitution of the purine salvage pathway in vitro and screening of inhibitors
50 L reactions were assembled in 384-well plates (Nunc TM 384-Well Clear Polystyrene) containing 100 mM Tris-HCl (pH 8.0), 100 mM KCl, 12 mM MgCl 2 ,1 mM DTT, 1 mM NAD + , 0.5 mM PRPP, and 0.5 mM NaH 2 PO 4 . The concentrations of all enzymes were initially set to 1 μM and later optimized to the following amounts, as described in the text: 15 nM PfADA, 1 μM PfHGXPRT, 62 nM PfPNP, and 62 nM SaIM-PDH. The change in fluorescence at 460 nm, after excitation at 340 nm, was monitored every 30s for 1 h. The mean of the first four fluorescence readings was used as a baseline and subtracted from subsequent measurements for each reaction. A standard curve was used to convert fluorescence values into concentration of NADH ( Figure S3 ). DMSO was used as a solvent for the library compounds in the screening assay. Control experiments showed that DMSO supplied at levels up to 10% did not substantially impair the reaction kinetics. Maximal velocities and area under the curve (AUC) values were derived from the progression curves using the R-script package grofit [52] . Z'-factors were calculated according to published methods [53] .
For screening, 35 μl of a common reaction mixture were distributed into each well and combined with 5 μl of individual compounds from the BioDesign library by Asinex (100 μM in 50% DMSO). Reactions were initiated by addition of 10 μl of 500 μM adenosine.

Reconstitution and optimization of the purine salvage pathway in vitro
The three enzymes of the purine salvage pathway from P. falciparum , along with the IMPDH from S. aureus , were expressed in Escherichia coli and purified by affinity chromatography ( Figure S1 ). PfHGXPRT requires activation in the presence of PRPP and Mg 2 + to enable enzymatic activity [49] . The activation time of PfHGXPRT was optimized to establish reproducible kinetics. Activation of the enzyme reached a plateau after 32 hours of activation in the presence of PRPP and Mg 2 + . The activated enzyme exhibited 9-fold increased activity compared to the non-activated enzyme ( Figure S2 ). Purified enzymes were individually assayed in vitro (see materials and method), and a common reaction buffer was identified, which supported activity of all four enzymes.
NADH was used as the readout of the activity of the complete pathway ( Figure 1 B ). NADH is highly fluorescent, with absorption and emission maxima at 340 and 460 nm, respectively, while NAD + is not fluorescent [54] . A linear correlation between concentration and fluorescence was measured at up to 80 nM NADH ( Figure S3 ). In vitro reconstitution of the purine salvage pathway required addition of four substrates: adenosine, inorganic phosphate, PRPP, and NAD + ( Figure 1 B ). The use of suitable concentrations of each substrate is essential for efficient detection of competitive inhibitors. Substrates supplied at levels that are too high will readily outcompete weak inhibitors, or inhibitors supplied at low concentrations [55 , 56] . Therefore, it is necessary to maintain substrate concentrations that are high enough to allow for synthesis of NADH, but that do not exceed ∼10-fold above their K M values. To select appropriate concentrations of individual substrates, the kinetic constants for each were determined ( Table 1 ). These kinetic constants are similar to values that were previously reported, demonstrating that the conditions used in the assay do not interfere with the enzymatic activities of the proteins. Based on these determinations, substrates were added to the reaction at concentrations equivalent to 3 -6 times their K m values, with the exception of PRPP which was supplied at 25-fold above K m (0.5 mM), to ensure adequate synthesis of NADH. The full pathway Step 1: IMPDH was then reconstituted using 1 μM of each enzyme. These initial conditions allowed for steady accumulation of NADH over time. Omission of a single enzyme from the reaction mixture led to a complete loss of fluorescence signal accumulation, demonstrating the absence of NADH synthesis ( Figure 1 C ). The assay exhibited excellent well-to-well reproducibility. NADH accumulation was analyzed by assessing the area under the curve as well as the reaction maximal velocities. The AUC of the complete reaction (4729 ± 72 μM min) exhibited a lower coefficient of variation (CV = 1.51%, n = 12) than that of the maximal velocity (5%, n = 12), and therefore represents a more robust descriptor of the kinetics of NADH synthesis. The experimental conditions also yielded excellent separation between the signals observed with the fully reconstituted pathway and reactions lacking a single enzyme ( Figure 1 C ). A calculated Z'-factor [53 , 57] of 0.96 (using the AUC) indicates that this assay is highly reliable for detection of inhibitors that completely inhibit at least one enzyme in the pathway. We hypothesized that this assay may also be useful for detection of partial inhibitors of any single enzyme in the reaction mixture. Assuming that inhibited and non-inhibited reactions exhibit equivalent CV values, a decrease of only 18% in NADH accumulation (i.e., the AUC) would still result in a Z'-factor equal to 0.5 ( Equation S1 ). Such a value is indicative of an assay that is very well suited for HTS, providing a separation of twelve standard deviation units between positive and negative values [53 , 57] .

Optimization of enzyme concentrations
The concentrations of enzymes were optimized to make the assay suitable for detecting partial inhibition of any of the four proteins in the pathway. The goals of this optimization were to i) match the kinetic rates of the individual steps in the pathway to each other, so that any step becomes rate limiting when partially inhibited, ii) maintain the levels of enzymes as low as possible, and iii) keep the concentration of NADH, used as a readout of pathway activity, within the range of reliable quantification. The concentration of enzymes can be difficult to optimize because the rate of NADH accumulation may not be linearly proportional to the amounts of the added enzymes, even if the rates of each step have been carefully tuned to one another. In other words, decreasing the concentration of an enzyme by 50% may not result in a 50% decrease in the rate of formation of the final product. This lack of direct proportionality between the amount of enzyme and the rate is explained by the accumulation of pathway intermediates within the K M range of the decreased enzyme. Therefore, reduction of the amount of a given enzyme leads to increased levels of the substrates for that enzyme, consequently increasing its saturation and compensating for the decreased amount of enzyme. Optimization of the concentration of each enzyme was achieved empirically as follows. All four enzymes were initially supplied in the reaction at a concentration of 1 μM, which produced a maximal rate of NADH synthesis and fluorescence signal The activity of the pathway was measured in a reaction mixture containing all four enzymes included at optimal concentrations as determined in Fig. 3 (Opt, n = 12), or with three enzymes supplied at optimal levels and the remaining enzyme reduced to 25% of the optimal amount (n = 12). Z'-factor values were calculated using the AUC at various timepoints (crosses with solid line).
( Figure 1 C ). Then, beginning with the final step of the pathway the concentration of enzyme ( i.e ., IMPDH) was decreased iteratively by half, until the AUC decreased by at least 18% compared to the previous concentration. This concentration was then maintained to optimize the concentration of the next enzyme in the pathway ( Figure 2 ). A single round using this approach was sufficient to pinpoint assay conditions in which a decrease by half in the concentration of one enzyme yielded at least an 18% reduction of the AUC during a 50 min reaction. An 18% reduction was chosen because such a decrease corresponds to a decrease by twelve standard deviations of a typical measurement of the AUC (see above and Equation S1 ). Optimization using this strategy resulted in input values ranging from 15 nM to 1 μM for each enzyme.

Detection of partial inhibition of any of the four enzymes of the purine salvage pathway
The optimized assay was tested for its ability to detect partial inhibition of any of the four enzymes of the pathway. To mimic partial inhibition, the concentration of a single enzyme was decreased by 75%, while the concentrations of the other three enzymes were maintained at optimal levels. This process was repeated for each of the four targeted proteins. Figure 3 shows that decreasing the amount of any one enzyme dramatically affected the kinetics of NADH synthesis. To identify the reaction time during which the AUCs exhibit the greatest separation between non-inhibited and inhibited conditions, Z'-factor values were calculated at various time points ( Figure 3 ). At a reaction time of 60 min, all the inhibition models yielded Z'-factors superior to 0.7, making this assay well suited for detection of partial inhibition ( > 75%) of any of the enzymes in the reaction mixture in a single run.

Pilot screening for inhibitors of the purine salvage pathway
A pilot screen was carried out using a sub-library of 384 representative compounds from the BioCore-based library, BioDesign by Asinex. Compounds were manually screened in batches of 24 during sixteen independent experiments (represented in different colors, Figure 4 A )  Compounds were screened in batches of 24, across 6-weeks (indicated by different colors). The hit threshold was defined independently for each batch as the mean minus three standard deviation cut-off (μ-3SD; dashed red line) [58] . B. Hit compounds and their targets. The targets of compounds exhibiting an AUC below the cutoff value were identified and Ki values were determined as described in Materials and Methods.
AUC of NADH synthesis detected in 60 min is shown for each reaction. These experiments were carried out over a period of six-weeks, using two different preparations of enzymes; a minor shift in the AUC data trend can be observed midway through the data set (between sets 8 and 9) due to the switch between enzyme preparations. Twenty compounds exhibited intrinsic fluorescence and could not be assayed. A threshold value for hit compounds was calculated for each batch using the mean AUC minus three standard deviations as previously described [58] . Individual targets of the hit compounds were identified by assaying against each enzyme individually. This pilot screen yielded five compounds with inhibitory effects against three of the four enzymes present in the reaction, with K i values ranging from 2-50 μM ( Figure 4 B ). No inhibitors of PfHGXPRT were identified, probably due to the limited size of the library used in the pilot screen. However, Figure 3 shows that reduction of HGXPRT in the reaction mixture decreases the activity of the overall pathway, indicating that the assay is suitable for identifying inhibitors of this enzyme as well.

Discussion
One limitation of conventional high-throughput screening assays is that they generally only screen against a single drug target. The one-target approach imposes a significant logistical (and in many cases financial) burden that considerably limits the potential for finding hit compounds during screening of chemical libraries. Incorporating multiple targets into a single HTS assay can substantially speed up the hit discovery rate because multiple targets are screened simultaneously. In addition, multi-target assays can be used with focused libraries if the targets in the assay exhibit similar substrates, which is often the case for enzymes belonging to the same metabolic pathway. We previously developed two multi-target HTS assays for screening of multiple drug targets in a single reaction. The first assay combined two targeted enzymatic sites and the second incorporated eight active sites at once [59 , 60] . The topology of the pathways used in those assays included no more than two consecutive enzymatic activities, with one system incorporating several enzymes working in parallel. The low number of consecutive steps greatly simplified optimization of the assays for detection of partial inhibition of any of the targets. The P. falciparum purine salvage pathway reconstituted in this assay includes four targets reacting in series from a single pathway. Our results show that such approach can be effectively utilized for detection of inhibitors targeting each individual enzyme in the pathway, albeit more attention may be required for optimization to adequately detect partial inhibition.

Declaration of Competing Interest
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.